There are some errors when I changed 'NUM_CLASSES'. about pytorch-deeplab

[speedinghzl]

You need to change the Line 181,
change
if not args.num_classes == 21 or not i_parts[1]=='layer5':
into
if args.num_classes == 21 or not i_parts[1]=='layer5':

[MarcokH]

thanks for your reply, it solved my problem,but i encounter another problem:
RuntimeError: cuda runtime error (59) : device-side assert triggered at /pytorch/torch/lib/THC/generic/THCTensorCopy.c:18
it happened in

Pytorch-Deeplab/train_msc.py

Line 230 in 795287e

loss100 = loss_calc(out, nd.zoom(labels, (1.0, out.size()[2]/d1, out.size()[3]/d2), order=0), args.gpu)

because of

Pytorch-Deeplab/train_msc.py

Line 107 in 795287e

label = Variable(label).cuda(gpu)

in fact,when i tried to solve the previous question ,i met the same problem.
could you give some advice to solve the problem? thanks!

[speedinghzl]

I suggest you use the train.py script, I don't update the multi-scale version.

[MarcokH]

Ok, I will try to run the train.py script. thanks again!